tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-07-20 01:27:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into compilade/mamba2

Browse Source
This commit is contained in:
Francis Couture-Harpin
2025-05-01 17:43:53 -04:00
parent c9ecf620d6 b1dd4d08e8
commit 35d06fac5a
462 changed files with 75597 additions and 48146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

195
ggml/include/ggml.h
View File

@ -393,8 +393,8 @@ extern "C" {
// precision
enum ggml_prec {
GGML_PREC_DEFAULT,
GGML_PREC_F32,
GGML_PREC_DEFAULT = 0, // stored as ggml_tensor.op_params, 0 by default
GGML_PREC_F32 = 10,
};
// model file types
@ -454,6 +454,7 @@ extern "C" {
GGML_OP_RMS_NORM,
GGML_OP_RMS_NORM_BACK,
GGML_OP_GROUP_NORM,
GGML_OP_L2_NORM,
GGML_OP_MUL_MAT,
GGML_OP_MUL_MAT_ID,
@ -480,6 +481,7 @@ extern "C" {
GGML_OP_CONV_TRANSPOSE_1D,
GGML_OP_IM2COL,
GGML_OP_IM2COL_BACK,
GGML_OP_CONV_2D_DW,
GGML_OP_CONV_TRANSPOSE_2D,
GGML_OP_POOL_1D,
GGML_OP_POOL_2D,
@ -502,20 +504,16 @@ extern "C" {
GGML_OP_ADD_REL_POS,
GGML_OP_RWKV_WKV6,
GGML_OP_GATED_LINEAR_ATTN,
GGML_OP_RWKV_WKV7,
GGML_OP_UNARY,
GGML_OP_MAP_UNARY,
GGML_OP_MAP_BINARY,
GGML_OP_MAP_CUSTOM1_F32,
GGML_OP_MAP_CUSTOM2_F32,
GGML_OP_MAP_CUSTOM3_F32,
GGML_OP_MAP_CUSTOM1,
GGML_OP_MAP_CUSTOM2,
GGML_OP_MAP_CUSTOM3,
GGML_OP_CUSTOM,
GGML_OP_CROSS_ENTROPY_LOSS,
GGML_OP_CROSS_ENTROPY_LOSS_BACK,
GGML_OP_OPT_STEP_ADAMW,
@ -680,6 +678,9 @@ extern "C" {
GGML_API bool ggml_is_contiguous_1(const struct ggml_tensor * tensor); // contiguous for dims >= 1
GGML_API bool ggml_is_contiguous_2(const struct ggml_tensor * tensor); // contiguous for dims >= 2
// true for tensor that is stored in memory as CxWxHxN and has been permuted to WxHxCxN
GGML_API bool ggml_is_contiguous_channels(const struct ggml_tensor * tensor);
GGML_API bool ggml_are_same_shape (const struct ggml_tensor * t0, const struct ggml_tensor * t1);
GGML_API bool ggml_are_same_stride(const struct ggml_tensor * t0, const struct ggml_tensor * t1);
@ -1095,6 +1096,18 @@ extern "C" {
int n_groups,
float eps);
// l2 normalize along rows
// used in rwkv v7
GGML_API struct ggml_tensor * ggml_l2_norm(
struct ggml_context * ctx,
struct ggml_tensor * a,
float eps);
GGML_API struct ggml_tensor * ggml_l2_norm_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a,
float eps);
// a - x
// b - dy
GGML_API struct ggml_tensor * ggml_rms_norm_back(
@ -1651,7 +1664,7 @@ extern "C" {
struct ggml_tensor * a,
struct ggml_tensor * b);
// depthwise
// depthwise (via im2col and mul_mat)
GGML_API struct ggml_tensor * ggml_conv_2d_dw(
struct ggml_context * ctx,
struct ggml_tensor * a, // convolution kernel
@ -1663,6 +1676,22 @@ extern "C" {
int d0, // dilation dimension 0
int d1); // dilation dimension 1
// Depthwise 2D convolution
// may be faster than ggml_conv_2d_dw, but not available in all backends
// a: KW KH 1 C convolution kernel
// b: W H C N input data
// res: W_out H_out C N
GGML_API struct ggml_tensor * ggml_conv_2d_dw_direct(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
int stride0,
int stride1,
int pad0,
int pad1,
int dilation0,
int dilation1);
GGML_API struct ggml_tensor * ggml_conv_transpose_2d_p0(
struct ggml_context * ctx,
struct ggml_tensor * a,
@ -1708,24 +1737,29 @@ extern "C" {
float p0,
float p1);
// nearest interpolate
enum ggml_scale_mode {
GGML_SCALE_MODE_NEAREST = 0,
GGML_SCALE_MODE_BILINEAR = 1,
};
// interpolate
// multiplies ne0 and ne1 by scale factor
// used in stable-diffusion
GGML_API struct ggml_tensor * ggml_upscale(
struct ggml_context * ctx,
struct ggml_tensor * a,
int scale_factor);
int scale_factor,
enum ggml_scale_mode mode);
// nearest interpolate
// nearest interpolate to specified dimensions
// used in tortoise.cpp
// interpolate
// interpolate scale to specified dimensions
GGML_API struct ggml_tensor * ggml_upscale_ext(
struct ggml_context * ctx,
struct ggml_tensor * a,
int ne0,
int ne1,
int ne2,
int ne3);
int ne3,
enum ggml_scale_mode mode);
// pad each dimension with zeros: [x, ..., x] -> [x, ..., x, 0, ..., 0]
GGML_API struct ggml_tensor * ggml_pad(
@ -1777,11 +1811,11 @@ extern "C" {
#define GGML_KQ_MASK_PAD 64
// q: [n_embd, n_batch, n_head, 1]
// k: [n_embd, n_kv, n_head_kv, 1]
// v: [n_embd, n_kv, n_head_kv, 1] !! not transposed !!
// mask: [n_kv, n_batch_pad, 1, 1] !! n_batch_pad = GGML_PAD(n_batch, GGML_KQ_MASK_PAD) !!
// res: [n_embd, n_head, n_batch, 1] !! permuted !!
// q: [n_embd_k, n_batch, n_head, 1]
// k: [n_embd_k, n_kv, n_head_kv, 1]
// v: [n_embd_v, n_kv, n_head_kv, 1] !! not transposed !!
// mask: [n_kv, n_batch_pad, 1, 1] !! n_batch_pad = GGML_PAD(n_batch, GGML_KQ_MASK_PAD) !!
// res: [n_embd_v, n_head, n_batch, 1] !! permuted !!
GGML_API struct ggml_tensor * ggml_flash_attn_ext(
struct ggml_context * ctx,
struct ggml_tensor * q,
@ -1891,85 +1925,18 @@ extern "C" {
struct ggml_tensor * state,
float scale);
GGML_API struct ggml_tensor * ggml_rwkv_wkv7(
struct ggml_context * ctx,
struct ggml_tensor * r,
struct ggml_tensor * w,
struct ggml_tensor * k,
struct ggml_tensor * v,
struct ggml_tensor * a,
struct ggml_tensor * b,
struct ggml_tensor * state);
// custom operators
typedef void (*ggml_unary_op_f32_t) (const int, float *, const float *);
typedef void (*ggml_binary_op_f32_t)(const int, float *, const float *, const float *);
typedef void (*ggml_custom1_op_f32_t)(struct ggml_tensor *, const struct ggml_tensor *);
typedef void (*ggml_custom2_op_f32_t)(struct ggml_tensor *, const struct ggml_tensor *, const struct ggml_tensor *);
typedef void (*ggml_custom3_op_f32_t)(struct ggml_tensor *, const struct ggml_tensor *, const struct ggml_tensor *, const struct ggml_tensor *);
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_unary_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
ggml_unary_op_f32_t fun),
"use ggml_map_custom1 instead");
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_unary_inplace_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
ggml_unary_op_f32_t fun),
"use ggml_map_custom1_inplace instead");
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_binary_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
ggml_binary_op_f32_t fun),
"use ggml_map_custom2 instead");
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_binary_inplace_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
ggml_binary_op_f32_t fun),
"use ggml_map_custom2_inplace instead");
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_custom1_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
ggml_custom1_op_f32_t fun),
"use ggml_map_custom1 instead");
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_custom1_inplace_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
ggml_custom1_op_f32_t fun),
"use ggml_map_custom1_inplace instead");
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_custom2_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
ggml_custom2_op_f32_t fun),
"use ggml_map_custom2 instead");
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_custom2_inplace_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
ggml_custom2_op_f32_t fun),
"use ggml_map_custom2_inplace instead");
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_custom3_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
struct ggml_tensor * c,
ggml_custom3_op_f32_t fun),
"use ggml_map_custom3 instead");
GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_map_custom3_inplace_f32(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b,
struct ggml_tensor * c,
ggml_custom3_op_f32_t fun),
"use ggml_map_custom3_inplace instead");
// custom operators v2
typedef void (*ggml_custom1_op_t)(struct ggml_tensor * dst , const struct ggml_tensor * a, int ith, int nth, void * userdata);
typedef void (*ggml_custom2_op_t)(struct ggml_tensor * dst , const struct ggml_tensor * a, const struct ggml_tensor * b, int ith, int nth, void * userdata);
typedef void (*ggml_custom3_op_t)(struct ggml_tensor * dst , const struct ggml_tensor * a, const struct ggml_tensor * b, const struct ggml_tensor * c, int ith, int nth, void * userdata);
@ -2025,6 +1992,30 @@ extern "C" {
int n_tasks,
void * userdata);
typedef void (*ggml_custom_op_t)(struct ggml_tensor * dst , int ith, int nth, void * userdata);
GGML_API struct ggml_tensor * ggml_custom_4d(
struct ggml_context * ctx,
enum ggml_type type,
int64_t ne0,
int64_t ne1,
int64_t ne2,
int64_t ne3,
struct ggml_tensor ** args,
int n_args,
ggml_custom_op_t fun,
int n_tasks,
void * userdata);
GGML_API struct ggml_tensor * ggml_custom_inplace(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor ** args,
int n_args,
ggml_custom_op_t fun,
int n_tasks,
void * userdata);
// loss function
GGML_API struct ggml_tensor * ggml_cross_entropy_loss(
@ -2141,7 +2132,11 @@ extern "C" {
# define GGML_RESTRICT
# endif
#else
# define GGML_RESTRICT restrict
# if defined (_MSC_VER) && (__STDC_VERSION__ < 201112L)
# define GGML_RESTRICT __restrict
# else
# define GGML_RESTRICT restrict
# endif
#endif
typedef void (*ggml_to_float_t) (const void * GGML_RESTRICT x, float * GGML_RESTRICT y, int64_t k);
typedef void (*ggml_from_float_t)(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);